Evolution of Crystallinity, Chain Mobility, and Crystallite Size during

The method is demonstrated by following the crystallization of natural rubber at −10 °C. At selected stages during the crystallization, a Bloch dec...
0 downloads 0 Views 136KB Size
8028

Macromolecules 2004, 37, 8028-8034

Evolution of Crystallinity, Chain Mobility, and Crystallite Size during Polymer Crystallization Johannes Leisen* and Haskell W. Beckham School of Polymer, Textile and Fiber Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0295

Mohammed A. Sharaf Department of Chemistry, Helwan University, Ain Helwan, Egypt Received May 20, 2004; Revised Manuscript Received August 9, 2004

ABSTRACT: A 1H NMR method is described for examining crystallinity, chain mobility, and crystallite size in real time during polymer crystallization. The method is demonstrated by following the crystallization of natural rubber at -10 °C. At selected stages during the crystallization, a Bloch decay was recorded along with Goldman-Shen decays for a series of mixing times using a filter duration to destroy the rigid-phase magnetization. The fast-relaxing components of the Bloch and Goldman-Shen decays correspond to the rigid/crystalline fraction of the sample and were fitted to Gaussian functions. The slow-relaxing components were obtained experimentally using the Goldman-Shen sequence measured with a short mixing time (1 ms). By combining the fitted Gaussian portions with the experimentally determined slow-relaxing components, the Bloch and Goldman-Shen decays could be analyzed across the entire crystallization process. The Bloch decays provided information on the amount and nature of the rigid fraction, while the Goldman-Shen spin-diffusion data provided information on domain size. The formation of noncrystalline rigid domains was observed for the initial stages of the crystallization process. Data clearly indicate lateral growth of crystallites with relatively uniform thickness during the primary crystallization process. Later stages of crystallization are characterized by increasing heterogeneity in crystalline density and perfection.

Introduction Crystallites in polymers exhibit a major influence on their bulk properties.1 Modifications to crystallinity can significantly alter mechanical or optical properties, for example. Thus, the thermal history of a polymer, in particular the details of its experience between Tg and Tm,2,3 greatly influences its end-use value. Hence, polymer crystallinity is characterized on a routine basis. Particularly useful analytical techniques allow the determination of both crystallite size and overall crystallinity in real time for kinetic studies. For mechanical properties above Tg, characterization of the interface between the mobile amorphous domains and a rigid phase such as a filler or a crystallite is also of interest.4 When the densities of amorphous and crystalline phases are known, crystallization processes in polymers may be followed by simple dilatometric techniques.5 Crystals growing in thin films may be readily observed with optical microscopy6 and light scattering techniques.7,8 These techniques can be used to determine the total amount and morphology of crystals, including formation kinetics. Molecular-level information on crystallite structure can be obtained by X-ray9,10 or electron diffraction methods.11 However, if kinetics are measured with X-ray techniques, it is often necessary to use expensive synchrotron radiation sources.12 Methods of nuclear magnetic resonance (NMR) have been used successfully to characterize the crystalline structure within many polymeric materials.13-23 In fact, relatively cost-effective low-resolution NMR scanners are used routinely to determine crystallinity in semi* Corresponding author: e-mail [email protected].

crystalline commodity polymers.24 The nucleus of choice is typically the 1H, whose sensitivity along with modern pulsed NMR instrumentation allows the measurement of % crystallinity in seconds or less. Hence, simple 1H NMR experiments may be used for time-resolved studies of polymer crystallization.13,14,20,22 These studies are based on measuring Bloch decays, which are then fitted to relaxation functions containing adjustable parameters that describe the relative amounts and nature of the underlying domains in the sample. When the crystalline, amorphous, and interphase domains are considered, up to eight fit parameters are required.19 This approach provides a wealth of information but is also associated with considerable ambiguity due to the use of so many adjustable parameters. Here we present a novel, very robust, and yet simple approach for investigating polymer crystallization by 1H NMR using fit functions composed of only three adjustable parameters.25 In addition to Bloch decays, spindiffusion data26 are also measured at a given stage of polymer crystallization. Thus, in addition to the total crystalline content, the crystallite size may be determined as a function of crystallization time. Even though it is not considered explicitly, our approach also provides insight into the nature of the interface between amorphous and crystalline domains. The method is demonstrated for crystallization of natural rubber at -10 °C. However, the experimental approach described here is anticipated to be of general use for the characterization of crystallization processes in protonated polymers. NMR Experiments For a semicrystalline polymer sample above its glass transition, the fraction of amorphous and crystalline

10.1021/ma0489974 CCC: $27.50 © 2004 American Chemical Society Published on Web 09/18/2004

Macromolecules, Vol. 37, No. 21, 2004

Polymer Crystallization 8029

phases may be simply obtained by deconvolution of a time-dependent Bloch decay, IBloch(t):

IBloch(t) ) wmobileFmobile(t) + wrigidFrigid(t)

(1)

{ ( )} t

/ T2,rigid

2

{ ( )}

IBloch(t) ) wrigid exp -

The fast-decaying component, Frigid(t), is attributed to rigid crystallites while the slow-relaxing component, Fmobile(t), is due to the mobile amorphous phase. Data can be normalized such that wrigid and wmobile represent molar fractions of mobile and rigid phases, respectively (i.e., wmobile + wrigid ) 1). Equation 1 neglects contributions from an interphase; evaluation of crystallinity in polymers based on this simplified two-phase model has been proven feasible.13,27,28 For kinetic studies, a suitable model function for IBloch(t) should describe the changes in the free-induction decays observed experimentally over the entire crystallization process and for a wide range of experimental conditions (e.g., different temperatures). On the time scale of the NMR experiment, chains within crystallites do not typically exhibit significant amounts of large-angle molecular motions so that 1H1H dipolar couplings govern the relaxation. Consequently, Frigid(t) can be modeled by a Gaussian function / is the decay constant: in which T2,rigid

Frigid(t) ) exp -

A more general approach is the combination of a theoretical function for Frigid(t) with experimental data measured for Fmobile(t):

(2)

The relaxation of protons in a well-defined crystalline lattice often appears as an FID with some sinusoidal character, the so-called Abragam function.29 This is certainly the case for semicrystalline polyethylene.17,30 For the natural rubber investigated here, there were no clear signs of this sinusoidal character in the FID (vide infra, cf. Figure 3). This signifies the absence of well-defined distances between individual protons in the rigid phase. Hence, the existence of small crystallites with large amounts of defects and even rigidified amorphous components is likely for natural rubber. The natural rubber studied here exhibits a calorimetric Tg at -66 °C (cooling rate of 10 °C/min); crystallization is easily observed from -40 to 10 °C. Over this temperature range, large changes occur for the amorphous phase, making it difficult to adequately model its behavior with a single Fmobile(t). At the lower temperatures from -40 to -20 °C, the molecular motion within the amorphous phase is sufficiently slow that 1H-1H dipolar couplings govern the relaxation. Consequently, Fmobile(t) can be modeled by a Gaussian function similar / as the decay to that shown in eq 2, but with T2,mobile constant. At the higher temperatures above -10 °C, large-angle molecular motions occur more rapidly and more frequently, leading to a reduction of any residual dipolar couplings. Consequently, chemically distinguishable nuclei contribute separately to the relaxation function (i.e., chemical shift resolution). A suitable model function may be found for Fmobile(t) under these circumstances, but it will be quite complex. It would also not be general. Because of the chemical shift influences, a model function would depend on the sample, the strength of the external magnetic field, and the excitation frequency. Hence, evaluation of IBloch(t) data measured at different temperatures would require the selection of temperature- and spectrometer-dependent fitting functions.

t

2

/ T2,rigid

+ wmobileIGS(t,tmf0) (3)

The experimental data, IGS(t,tmf0), are obtained by selective measurement of slow-relaxing magnetization using a Goldman-Shen (GS) pulse sequence with a filter to destroy the rigid-phase magnetization and a mixing time (tm) that is short enough to prevent significant transfer of magnetization back into the rigid phase.31 This GS decay corresponds to mobile-phase magnetization and can be attributed to the amorphous fraction within a semicrystalline polymer. At lower temperatures, where a Gaussian decay function like that shown in eq 2 may be applicable for both amor/ phous- and crystalline-phase magnetization, the T2,rigid / of eq 3 must be significantly smaller than the T2,mobile to distinguish between phases. In fact, this condition is essential for any model function chosen for IBloch(t) and also for using the GS sequence to selectively measure Fmobile(t). The Goldman-Shen and other T2-based filter sequences such as the dipolar filter have been used for the characterization of polymer morphologies.26 Following application of the filter, the selected mobile-phase magnetization is exposed to a variable-duration mixing time (tm). For very short mixing times, the detected magnetization corresponds to the mobile phase. For longer mixing times, the selected mobile-phase magnetization will transfer to rigid domains via dipolarcoupling-mediated spin diffusion. The amount of mobilephase magnetization, mmobile(tm), transferred to the rigid domains for a given mixing time, mrigid(tm), may be approximated by analyzing the respective GS decay using an equation analogous to eq 3:

{ ( )}

IGS(t,tm) ) mrigid(tm) exp -

t

/ T2,rigid

2

+

mmobile(tm) IGS(t,tmf0) (4) For short mixing times (i.e., tm f 0), mrigid(tm) ) 0 and mmobile(tm) ) 1. As the mixing time is increased, mrigid(tm) increases and mmobile(tm) decreases until complete equilibration of magnetization occurs. At sufficiently long mixing times (i.e., as tm f ∞), IGS(t,tm) assumes the shape of IBloch(t) but with lower overall intensity, mmobile(tm) ) wmobile and mrigid(tm) ) wrigid. The spin diffusion data are analyzed by fitting eq 4 to determine mmobile(tm), which is then plotted as a function of xtm. Following the approach of Mellinger et al.,32 the initial slope of this plot is extrapolated to mmobile(tm) ) 0 to find t/m, which is used to calculate the size of the mobile domains. This calculation requires spin-diffusion constants, which can be obtained from the T2 relaxation time using an established calibration curve.32 The determination of polymer crystallinity using eq 3 requires two separate NMR experiments to measure: (1) a Bloch decay and (2) a Goldman-Shen decay. Depending on the T1 relaxation times and number of data accumulations, these two experiments can be performed within several tens of seconds. If domain sizes are to be determined for a single stage during an ongoing crystallization process, Goldman-Shen data

8030

Leisen et al.

Macromolecules, Vol. 37, No. 21, 2004

Figure 1. Sequence of experiments used to measure crystallinity, chain mobility, and crystallite size during a single stage of an ongoing polymer crystallization process. A Bloch decay is followed by Goldman-Shen decays with different mixing times. The filter duration on the Goldman-Shen sequence is set to destroy the rigid-phase magnetization; the nth GoldmanShen decay will be similar in shape to the Bloch decay, but with lower overall intensity. The measurement time for this entire data matrix must be very small compared to the total crystallization time.

need to be recorded for multiple mixing times. This can be typically achieved in several minutes. This time frame is sufficiently short to neglect changes in the nature and amount of growing crystallites for crystallization processes that occur over several hours. The entire experimental sequence is shown in Figure 1. Analysis of spin-diffusion data based on eq 4 depends / , which is the third adjustable fit parameter on T2,rigid (in addition to the mobile- and rigid-phase magnetization). This parameter reveals additional and perhaps unanticipated information on semicrystalline polymer morphology. In a sample without an interphase between its crystalline and amorphous domains, crystallinephase magnetization will be characterized by a single / that should not change as a function of value of T2,rigid tm. A more realistic scenario is that of a sizable interphase. For a polymer between Tg and Tm, chains within the interphase will exhibit mobilities intermediate to mobilities of chains in crystalline and amorphous regions. Consequently, the T/2 values for an interphase should be greater than those measured for the crystallites. In a spin-diffusion experiment, mobile-phase magnetization must travel through this interphase on its way to the rigid crystalline phase. Since the model function used to fit the GS decays (eq 4) only contains / , the magnitude of this a single T/2 parameter, T2,rigid parameter should reflect the relative fraction of interphase in the sample (cf. Figure 2). If a significant interphase exists, it will be most obvious for short / that is mixing times and characterized by a T2,rigid / greater than the T2,rigid measured for a well-crystallized sample. For longer mixing times, as most of the magnetization has passed through this interphase, the / should be more representative of the rigid/ T2,rigid crystalline phase. If the relative fraction of interphase to crystalline phase changes as a function of crystallization time, this should be observed as changes in plots / vs mixing time. of T2,rigid Experimental Section The 1H NMR experiments were conducted on a Bruker DSX 300 spectrometer (Bruker Biospin, Rheinstetten, Germany) at

Figure 2. Magnetization pathway in a polymer with an interphase during a spin-diffusion experiment (left column, depicted as relative concentrations) along with the corresponding molecular mobilities (right column, represented by T/2 / values). Note the existence of a distribution of T2,rigid values after spin diffusion carries magnetization into the interphase. / Fitting data with eq 4 returns a single value for T2,rigid , in which information is embodied on the relative fractions of interphase and crystalline phases. a magnetic field strength of 7.1 T. A static double-resonance probe with a 7-mm coil was used. The temperature within the probe was controlled using the variable-temperature accessory of the spectrometer. Prior to the NMR measurements, the natural rubber sample was held at 70 °C for about 30 min and then quenched to ensure that no crystallites were present at the beginning of the experiment. Crystallization of the pure natural rubber at -10 °C was observed by 1H NMR spectroscopy for 24 h. Bloch decays were measured and followed immediately by Goldman-Shen experiments (cf. Figure 1) for 16 different mixing times from 1 to 1500 ms. The time required to measure these 17 different decays, referred to here as a “stage” of the crystallization, was about 18 min. Crystallization occurred most rapidly within the first 10 h of the experiment (see Figure 4); during this time, the crystallinity reached nearly 30%, and 30 stages were measured. Thus, during a single stage, the change in the total crystallinity of the sample was less than 1%. Even so, single-stage data sets were not recorded with consecutively increasing values of tm but with randomly ordered mixing times to avoid errors in data interpretation caused by systematic changes in the sample. For both the Goldman-Shen and Bloch decays, a repetition delay of 4 s was used, and 512 complex data points were acquired using a dwell time of 1 µs. The Goldman-Shen sequence consisted of three π/2 pulses with a length of 3 µs. The time interval between the first two pulses, which acts as

Macromolecules, Vol. 37, No. 21, 2004

Polymer Crystallization 8031

Results and Discussion

Figure 3. Experimental 1H NMR free-induction decays for semicrystalline natural rubber after crystallization for 24 h at -10 °C: Bloch decay (×) and Goldman-Shen decays recorded using a 30-µs filter duration (marked by vertical dashed line) to destroy rigid-phase magnetization followed by these mixing times, tm: 1 ms (+), 5 ms (2), 8 ms (1), 24 ms (9), and 50 ms ([). Solid lines are fits to eq 3 (Bloch decay: ×) or eq 4 (Goldman-Shen decays: 2, 1, 9, [) using the data set IGS(t,tm ) 1 ms) (+) as an approximation for IGS(t,tm f 0). For each data point, errors are less than the symbol sizes.

Figure 4. Fraction of rigid components (i.e., crystallites) vs time for the crystallization of natural rubber at -10 °C. The sigmoidal curve is a fit to the initial data using the Avrami equation.33,34 The slower but continuing increase in crystallinity after about 10 h is indicative of secondary crystallization processes. Experimental error is (0.2% for each data point. the filter, was set to 30 µs. This filter time is sufficiently long that rigid-phase magnetization is completely destroyed (cf. Figure 3). For the Bloch decay, a somewhat shorter pulse length of 1 µs was used, primarily to allow the use of a receiver gain setting that was also appropriate for the subsequent collection of Goldman-Shen data. All data were recorded and stored within a single data matrix, which was then transferred to a personal computer and analyzed using the IGOR-Pro software package (Wavemetrics, Lake Oswego, OR). Steps of this analysis included sorting of free-induction decays according to mixing times, an automatic phase correction to shift spectral intensities into the real part, which were then fitted to either eq 3 (Bloch) or eq 4 (GS). The Goldman-Shen decay with the shortest mixing time (1 ms) was used as IGS(t,tmf0). The pulse lengths and deadtime duration (5 µs) were taken into account by fitting the time-domain data using a time axis that accurately reflected the onset of data sampling after the center of the last rf pulse. Thus, relaxation during the dead time should not significantly affect the results. The use of solid echoes for data detection minimizes dead-time relaxation effects, which is more important if frequency-domain data were being analyzed. However, such sequences require longer phase cycles which can lead to prolonged measuring times for samples with sufficient signal that the number of scans are the minimum required to advance through one phase cycle. For kinetic studies, prolonged measuring times are generally undesirable.

Crystallinity. Bloch decays were measured about three times an hour for nearly 24 h while a sample of natural rubber was crystallizing at -10 °C. As described above and shown in Figure 1, these Bloch decays were immediately followed by a series of Goldman-Shen decays. A representative portion of a data set is shown in Figure 3 for the crystallization stage at 24 h. The solid lines demonstrate the quality of the fits to eq 3 (Bloch decay) or eq 4 (Goldman-Shen decays) and the validity of this approach for analyzing kinetic crystallization data. There is a slight deviation between the experimental Bloch decay and its fit, most likely related to the use of the Gaussian (eq 2) instead of the Abragam function29 to describe the rigid component. The fraction of rigid components (wrigid × 100%), obtained by fitting the experimental Bloch decays to eq 3, is plotted as a function of crystallization time in Figure 4. The detected increase in rigid components clearly reflects the growing crystallinity. The data in Figure 4 reveal the presence of both primary and secondary crystallization events as often observed in polymers.13 Primary crystallization occurs up to about 10 h. It can be described by the Avrami equation33,34 with a rate constant, k ) 4.9 × 10-5 s-1, and an Avrami exponent, β ) 3.0 (see dashed sigmoidal curve in Figure 4). These values compare well to those reported for crystallization of cis-1,4-polybutadiene at a temperature above Tg similar to that used in this study (55 K above Tg).13 After 10 h, crystallinity continues to grow but at a much reduced rate; this region is attributed to secondary crystallization. Note the initial rigid-component fraction in Figure 4 / is not zero, but about 2.3%. The associated T2,rigid values obtained from the fits of the data from these initial stages of the crystallization are in the range 100200 µs, which are significantly larger than the values (12-18 µs) determined for the rigid-component fraction in a well-crystallized sample. Hence, the rigid-component fraction extracted through eq 3 may not necessarily correspond to crystallites in the early stages of the crystallization. Dynamic heterogeneities, characterized by a distribution of correlation times, most certainly exist in the precrystal amorphous matrix. The more rigid components in this matrix possibly represent nascent crystal nuclei composed of locally aligned chains.35 At later stages of the crystallization process, the rigid-component fraction determined from eq 3 largely represents polymer crystallites. Domain Sizes. At each stage of the crystallization process, a series of Goldman-Shen decays were recorded and analyzed with eq 4 to yield the mobile-phase magnetization, mmobile(tm), for different mixing times (tm). Thus, data like those shown in Figure 5 were generated for each stage of the crystallization. The observed mmobile(tm) values decreased with increasing tm until a plateau value was reached; this plateau value of mmobile(tm) is experimentally very close to the mobilecomponent fraction, wmobile, obtained from analysis of the Bloch decay using eq 3. This is strong confirmation that virtually all crystallites are captured by the spindiffusion process. The average domain size was calculated from these data exactly as outlined by Mellinger et al.32 The initial slope was extrapolated to mmobile(tm) ) 0 to yield the intercept with the abscissa, t/m. These tm/ values were

8032

Leisen et al.

Macromolecules, Vol. 37, No. 21, 2004

Figure 5. Mobile-phase magnetization, mmobile(tm), at a given crystallization stage for natural rubber crystallizing at -10 °C. The crystallinity of the sample at the time of these particular measurements was 34%. The mmobile(tm) values decrease to a plateau value, which is very close to the mobilecomponent fraction obtained by fitting a Bloch decay to eq 3 (wmobile, shown as the horizontal dashed line). The initial slope of this plot is extrapolated to mmobile(tm) ) 0 (shown as diagonal solid line) and the intercept with the abscissa, (t/m)0.5, is used to determine the mobile-component domain size.

determined at each stage of the crystallization process and used to calculate the mobile-component domain size, dmobile:32

dmobile )

2 Dmobilet/m x xπ

(5)

where  is a constant describing the dimensionality of the diffusion process and Dmobile is the mobile-phase spin-diffusion coefficient. For a lamellar structure, as is likely for a polymer crystallite,  ) 1. The spindiffusion coefficient, Dmobile, was determined from the spin-spin relaxation time [obtained by approximating the IGS(t,tmf0) decay by an exponential] using a calibration curve published by Mellinger et al.32 Since spin diffusion occurs with significantly larger diffusion coefficients in rigid domains, the rate-determining step for equilibration of magnetization is spin diffusion through the mobile domains. Therefore, the domain sizes were estimated here using mobile-phase spin-diffusion coefficients only, while rigid-phase spin-diffusion coefficients were ignored. This assumption may lead to some inaccuracies in the absolute domain sizes. However, since Dmobile did not change significantly throughout the entire crystallization process, trends in domain sizes are expected to be accurate. Spin-lattice relaxation occurs during the mixing time, but with such large time constants (T1) that it was not necessary to correct for this effect. For example, the overall sample T1 is 850 ms at the beginning of the crystallization process and 920 ms after 24 h. Since only mmobile(tm) values for tm e 25 ms were used to determine t/m (e.g., the initial slope in mmobile(tm) vs t0.5 m plots), T1 relaxation during this period is negligible. Assuming identical densities for both mobile and rigid domains, the size of the detected domains, dmobile, is related to the volume defined by a long period, d ) drigid + dmobile:  ) (dmobile + drigid)wmobile dmobile

(6)

If the rigid domains are assumed to be lamellar crystals,  ) 1 and their size (drigid) can be calculated from the mobile domain size using a simple relationship:

drigid ) dmobile

wrigid wmobile

(7)

Figure 6. Sizes of mobile (O) and rigid domains (0) as a function of crystallization time for natural rubber at -10 °C. Calculated experimental error for mobile-domain data is (0.3 nm and (0.6 nm for rigid-domain data.

Values for wrigid/wmobile can be directly obtained from the data shown in Figure 4. Mobile- and rigid-domain sizes were thus obtained for each stage during the crystallization and are shown in Figure 6 as a function of crystallization time. Domain sizes were not obtained for the very early stages of the crystallization process ( 4 ms), the T2,rigid

Polymer Crystallization 8033

values fall below 20 µs, which is characteristic of chains within crystalline domains. The abrupt change from / / ) 60 µs (interphase) to T2,rigid < 20 µs (crystalT2,rigid lite) results from using a fit function with a single relaxation constant (eq 4). If the fraction of transferred magnetization is dominated by crystalline domains, a / value characterisleast-squares fit will return a T2,rigid tic of chains within crystallites. At intermediate and later stages of the crystallization / process, all values of T2,rigid are